Parrot is an imitation learning approach to automatically learn cache access patterns by leveraging Belady’s optimal policy. Belady’s optimal policy is an oracle policy that computes the theoretically optimal cache eviction decision based on knowledge of future cache accesses, which Parrot approximates with a policy that only conditions on the past accesses.
Source: An Imitation Learning Approach for Cache ReplacementPaper | Code | Results | Date | Stars |
---|
Task | Papers | Share |
---|---|---|
Image Generation | 1 | 10.00% |
Reinforcement Learning (RL) | 1 | 10.00% |
Text-to-Image Generation | 1 | 10.00% |
Image-to-Image Translation | 1 | 10.00% |
Pedestrian Detection | 1 | 10.00% |
Instruction Following | 1 | 10.00% |
Classification | 1 | 10.00% |
Image Classification | 1 | 10.00% |
Ethics | 1 | 10.00% |
Component | Type |
|
---|---|---|
Linear Layer
|
Feedforward Networks | |
LSTM
|
Recurrent Neural Networks | |
Scaled Dot-Product Attention
|
Attention Mechanisms | |
Softmax
|
Output Functions |